JPH10234868A - Systems and methods for data compression and non-linear sampling for use in implantable and battery-powered devices and the like - Google Patents

Abstract

(57) Abstract: A system and method for digitizing and compressing an analog signal generated by a battery-operated device such as an implantable pacemaker, wherein the compression ratio is improved, the signal processing amount is reduced, and the power is reduced. Reduce consumption. A time-varying threshold signal is generated, and each time a sample clock is generated, the analog signal is compared with the threshold signal. If the analog signal does not exceed the threshold signal, sampling is skipped without taking a sample. Perform non-linear sampling. The threshold signal can be either an analog level signal or a plurality of discrete level signals. A compression process is performed on the non-linearly sampled data. At this time, both the amplitude value data and the data representing the number of samples that have been forgotten are compressed. The data thus obtained is subsequently decompressed. In a pacemaker, the compressed data
Data indicating the occurrence time and type of the marker channel event is added.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

TECHNICAL FIELD The present invention relates to an analog signal,
A system for compressing and sampling a physiological signal such as an electrocardiographic signal, etc., which can perform efficient data compression with a relatively small amount of data loss and a very small amount of data processing itself. As for the system.

[0002]

BACKGROUND OF THE INVENTION Over the past decade or so, significant advances in technology have led to the integration of processing circuitry and memory for storing digitized physiological signal episodes in battery-powered implantable devices. Is now possible. For example, in the case of a pacemaker, it is possible to detect and digitize an endocardial signal, store the digital data, and upload the stored data to an extracorporeal device later. ing. Further, in the pacemaker system configured as described above, it is possible not only to collect electrocardiographic signals as data but also to detect and digitize a sensor signal such as a blood pressure sensor signal and store it. The sampling of these signals is typically performed continuously at a high sampling rate, which is so for capturing large amplitude, high frequency components that occur, for example, when the heart actually contracts. This is because a proper sampling is required. in this way,
A specific example of a system that has already been put into practical use and has a function of detecting, digitizing, and storing physiological signals is a pacemaker / cardio having functions of three types of devices, a pacemaker, a cardioverter, and a defibrillator. Barter / defibrillator device (PCD),
There are implantable blood flow monitoring devices and the like.

[0003]

In these systems, and particularly in a battery-powered implantable system, a process for digitizing and storing data, and a process for storing stored data in an extracorporeal receiving apparatus. Difficulty related to the process of sending data to is increasing. The reason for this is that a sufficient data processing time must be secured for performing these processes, and a large amount of power is consumed for that purpose. In general, the memory capacity of the system is limited, but even if the memory capacity is not a significant constraint, some kind of data compression is required to reduce the power consumption associated with data processing and execution of telemetry. It is desirable to use a method. By adopting the data compression method, the amount of signal information that can be stored in the memory mounted on the system can be increased. This is described in U.S. patent application Ser. No. 08 / 561,838, filed Nov. 22, 1995 (case number: P-3432,
Title of the invention: systems and methods for compression of digital signals in battery-powered implantable devices).

The amount of available resources, both for signal processing and for battery power, is naturally limited, so that for the compression schemes used in implantable systems such as those described above, the amount of information loss is limited. It is important to increase the data compression ratio as much as possible while staying within the allowable range. That is, in the environment of an implantable system as described above, a lossy compression method (a compression method in which some information is lost during compression and the original signal cannot be completely restored) is employed to reduce the loss. It is conceivable that a high compression ratio can be obtained instead of the occurrence of the signal distortion.

[0005]

SUMMARY OF THE INVENTION The present invention has been made in view of the above technical background, and a system and method according to the present invention provide a conventional system for compressing a digitized physiological signal. The lossless compression method (a compression method in which information is not lost during compression and the original signal can be completely restored) is improved. This conventional lossless compression method achieves a compression ratio of about 2: 1. This is a compression method. The compression method according to the present invention is an improvement and enhancement of the conventional compression method, in which a larger compression ratio can be achieved instead of accepting the occurrence of some signal distortion. It can be said that this is a lossy compression method. The basic concept of the lossy compression scheme according to the present invention is to allow some signal distortion in the less important part of the signal, i.e., the part that contains only a small amount of information,
In a portion of high importance such as a portion of the actual electrocardiogram signal where the QRS complex is generated, the signal distortion is suppressed as small as possible. As a result of such an improvement and enhancement of the conventional compression scheme, samples of the signal are discarded according to the activity of the signal, i.e., the less important parts of the signal are discarded. The sample is to be discarded. In addition, the sampling interval is not constant. Furthermore, in this non-linear sampling method, among the acquired samples, the samples representing the data important part of the signal are:
What is necessary is just to compress and store by the conventional lossless compression method based on which the present invention is based, so that the period during which the sample representing the data important part is captured, that is, the part of the original signal, Can be accurately restored.

In addition, the system of the present invention, when applied to a pacemaker environment, provides a scheme for generating an annotated compressed signal by adding pacemaker markers to the compressed data stream. is there.

[0007] The invention also provides, in another of its aspects,
The present invention provides a system and method for performing nonlinear sampling by hardware before actually performing digital sampling with A / D conversion. . As described below according to the preferred embodiment,
A high quality ECG or EGM signal can be obtained with only 20% to 25% of the number of samples required in a conventional system without sample discard. This not only reduces the amount of data that must be stored, but also reduces battery power consumption. This is because the number of samples that the implantable device has to process, ie the number of samples that need to operate on it is, for example, 200 / sec, but on average about 50 / sec Because it becomes. The present invention also provides a subsequent processing method which is performed after performing the above-described nonlinear sampling and further performing the A / D conversion. This is a method for further compressing the data to be compressed.

[0008] It is an object of the present invention to provide a system for use in devices with limited power consumption, such as battery-powered implantable devices, which have improved data sampling and compression processes to provide compression. It is to optimize the trade-off between the ratio and the amount of signal loss, and thereby to optimize the processing and signal storage capacity for signal digitization of the system. In order to achieve this overall goal, an improved system is provided in both sampling and data compression. Still another object of the present invention is to provide a non-linear sampling method that minimizes the amount of signal loss. In this sampling method, a sampling is performed without capturing a sample representing a part of the input signal that contains little data, On the other hand, with respect to a data important part of the input signal, the maximum number of samples is taken in so as to minimize the signal distortion. This non-linear sampling method reduces the effective sampling rate. The acquired samples are subjected to linear compression. In this way, by combining nonlinear sampling and linear compression, the required data processing amount and required storage capacity can be greatly reduced as a whole. I have.

In one embodiment of the present invention, non-linear sampling is performed after initial sampling and A / D conversion, thereby discarding samples with a small amount of information and effectively compressing the data. Can be done. In another preferred embodiment, non-linear sampling is performed prior to A / D conversion, thereby further reducing the power consumption required for data processing. Also, when the signal is decompressed by combining the digital data representing the acquired samples with the data representing the number of samples skipped without being acquired, and linearly compressing both of the two types of data, the signal is decompressed. The original signal is restored by minimizing the data loss caused by missing the sample.

[0010] Yet another feature of the system of the present invention is a feature applicable to implantable pacemakers, which includes marker channel information representing pacemaker events, such as the delivery of pacing pulses. Then, it is distributed and compressed and stored in the sample data.

[0011]

Embodiments of the present invention will now be described with reference to the drawings.

First, FIG. 1A will be described. The figure shows
1 schematically illustrates the environment of the apparatus and method according to the present invention. The present invention can be applied to an implantable device such as a pacemaker 30, and the figure shows a state in which the pacemaker 30 is implanted in a patient's body. A lead 31 is connected to the pacemaker 30, and this lead 3
1 is drawn inside the patient's heart. Lead 31
Has one or more electrodes at its tip,
Through the electrodes of the lead 31, sending out a stimulation pulse,
An endocardial signal or an epicardial signal is detected. As is well known in the field of pacemakers, this pacemaker
It has a function of receiving and digitizing the detection signal and storing the digitized signal in a memory. The signal stored in the memory is transmitted to the extracorporeal device 37 later. As another operation method different from the above, it is also possible to directly download a digitized detection signal to a programming device which is an extracorporeal device. Similarly, when a sensor is attached to the lead 31 or a sensor is built in the pacemaker 30, a signal transmitted by the sensor can be digitized and stored. As shown, the transmitting / receiving device 37 is, for example, a programming device generally used in the field of pacemakers. When the programming device receives the data from the pacemaker, it can transfer the data to processor 38, which can output the data to input / output device 39, all of which can be in a known manner. Done. As is well known, an implantable device such as a pacemaker has a limited data processing capability, so that it has been difficult to handle an enormous amount of data. In addition, the limit of the data processing capacity was due to restrictions on power consumption and memory capacity.

FIG. 1A further shows the body surface-mounted device 32 worn on a patient's arm by a wearing band. This type of device 32 is, for example,
Used in combination with the electrodes 33 to detect the ECG of the patient. Like the implantable device, the battery-operated device 32 can also download the collected data to another device, further process the data in that device, and output the processed data from that device. It is like that. Such battery-powered body-worn devices are also subject to the limitations on available processing and memory capacity described above. still,
Although FIG. 1A illustrates two types of devices, the present invention is not limited to any particular type of device.

Referring now to FIG. 1B, there is shown a simplified block diagram illustrating the major components of an implantable device 30 or a body surface mounted device 32 according to the present invention. is there. Battery 25 supplies power to one or more sensors 26, which have leads 31-33. The sensor block 26 may further include amplifiers and digitizing circuits of a general configuration. The digitized data is transmitted to a processor block 34, which incorporates a microprocessor in a suitable configuration. The processing steps included in the present invention can be executed by appropriately combining dedicated hardware and software. The microprocessor can communicate with a memory (that is, a storage device) 35, and the memory 35 can communicate with a transmission / reception circuit 36. Thus, it is possible to transmit data to an external device and to receive data and instructions from the external device.

Next, FIG. 2 will be described. Shown in the figure is a simplified flowchart illustrating the major steps performed in data compression, transmission, and decompression according to the present invention. In step 40, the signal (which is, for example, a physiological signal such as an EKG signal) is
Pick up in the form of an analog signal. Next step 4
At 1, sampling of this analog signal and conversion from analog data to digital data (A / D conversion) are performed. As will be described in detail below, the procedure for performing these operations may be to first perform A / D conversion and then discard some of the samples so that non-linear sampling is performed. Yes, and let it go without capturing some of the samples,
A / D conversion may be performed on only the acquired sample. The digital data thus captured is stored in the buffer memory as shown in step 42. In step 44, the digital data is compressed using any one of several methods according to the invention, and in step 45 the compressed data is stored. In step 46, a command for data transmission is received from an external device,
This command indicates that the external device is ready to receive the stored data. In step 47,
Compressed data is transmitted, and this data transmission is performed by a general method well known in the field of pacemakers.
In a subsequent step 48, the data is decompressed in an external device, whereby the original signal is restored in digital form. In the following step 49, the restored signal data is stored or displayed on a display.

Next, FIG. 3 will be described. Shown in the figure is a flowchart illustrating the steps performed in the lossless compression scheme disclosed in the earlier filed US patent application Ser. No. 08 / 561,838. Generally, this compression method achieves a compression ratio of about 2: 1 and requires very little data processing. Therefore, this compression method is suitable for use in various devices having limited electric energy such as an implantable device. The basic technique of this compression algorithm is that the value of the previous sample (LastSample) and the value of the current sample (S
delta value (Delta) which is a difference value between
a, including the size and the sign) with the minimum necessary number of bits. Since the value of the 0th sample is unknown, it is assumed that the value corresponds to a 0 volt input signal, and the value is
It is used as an integration constant when restoring a signal. More specifically, first, as shown in step 51, the value of the variable LastSample is set to “0”. Steps 53-57 form a loop for assembling one data block containing a plurality of delta values. The optimal size of this data block is such that it contains between 5 and 12 samples. In step 52, the value of the variable #Samples is set to “0”, and the value of the variable MaxDelta is set to “0”. In step 53, the routine waits for a sample clock to occur. The sample clock is transmitted at a predetermined constant clock rate. When the sample clock is received, the operations shown in step 54 are performed. That is,
One delta value (Delta) is assigned to a variable #Sample.
s is stored in the buffer (Buffer [#Samples}) corresponding to the current number of samples. Subsequently, at step 55, the absolute value (| Delta |) of the delta value is stored in the variable Max currently stored.
It is determined whether or not the value is greater than the value of Delta. If so, the value of the variable MaxDelta is reset equal to its absolute value; otherwise, the value of the variable MaxDelta is kept unchanged.
In step 57, the value of variable #Samples is
determines whether the value of the buffer is equal to the value of bufferSize (this variable represents the buffer size), and determines whether the above processing has been completed for all the samples to be included in the current data block. Is what you do. If the result of this determination is negative (No), the routine returns to step 53 and executes the loop described above again. On the other hand, if the decision result in the step 57 is affirmative (Yes), the routine continues to a step 58, in which the value of the variable NBits is determined. The variable NBits indicates the number of significant bits of the value of the variable MaxDelta, and the number of significant bits is, for example, 3 bits or 4 bits. The variables (parameters) NBits can be determined using hardware equipped with a priority encoder, or can be determined by software. In the case of using software, for example, the variable MaxDelete
What is necessary is just to count the number of times of right shift required to set the value of a to “0”. Next next step 59
Then, the data to be included in the current data block is written to the main memory of the device for storing signals. When writing the data, first, the value of the number of significant bits is written as a block header, and subsequently, the delta values assembled as described above are written one after another. Also, when writing the delta value,
From the least significant bit of each delta value, the variable NBits
Only bits up to the number represented by the value of are used, and these bits are written with a sign bit indicating whether the delta value is positive or negative.

The five specific electrocardiographic signals were processed according to the above-described compression method, and the results were analyzed. As a result, all of the five types of signals were 2: 1 on average.
It was found that compression was performed at the above compression ratio. The average compression ratio of these five signals was about 55%. The reconstruction of the signal is also performed for each individual data block, and the reconstruction of this signal is disclosed in detail in US patent application Ser. No. 08 / 561,838. That is, first, the variable NBits
, Followed by the reconstruction of all delta values contained in the data block, where the sign bit is added to the least significant bit up to NBits bits of each delta value. Is read. Then, the original signal is restored by adding the delta value to the value of the preceding sample.

Next, FIG. 4 will be described. FIG.
7 shows a lossy compression algorithm capable of obtaining a higher compression ratio than the lossless compression algorithm of FIG. Several lossless signal compression schemes obtained by improving the lossless signal compression scheme shown in FIG. 3 have already been proposed.
When experiments were conducted on these compression methods, the improvement in compression ratio obtained by these compression methods was only about 10%, and even in order to achieve such an increase in compression ratio, processing resources and memory resources Has to be significantly increased. Therefore, in order to greatly improve the compression ratio, it is necessary to allow some distortion in the restored signal. However, in order for the reconstructed signal to have a quality high enough to be used for medical purposes, it is necessary to minimize the distortion. The sampling rate of a physiological signal, such as an ECG, is generally quite high because it must be done to accurately capture the fastest-varying components in the signal. The sampling rate of the respiratory signal must be high in order to accurately capture the signal when exercising, but the sampling rate may be reduced during rest or sleep. The sampling rate of the ECG signal is QR
In order to capture the S group, it must be at least 200 Hz, but it is possible to lower the rate if only the P wave and the T wave need to be captured because the waveforms are relatively large and spread. It is. In addition, during the period from the end of one QRS group to the start of the next QRS group, even if the sampling rate is further reduced, the amount of information loss does not increase so much. Given these considerations, an algorithm that adaptively adjusts the sampling rate to an effective sampling rate for each data block according to the activity of the signal in each data block. It can be appreciated that the adoption of can improve the lossless compression method described above. In this case, the lower the activity of the signal in a data block, the more samples need to be discarded, while the data blocks with high signal activity do not discard any samples, Alternatively, only a small amount may be discarded. By doing so, it is possible to obtain a lossy compression method in which distortion is very small in a portion of the analog signal having high signal activity.

Next, the flowchart of FIG. 4 will be described. As can be seen from the illustrated flow chart, the first part of the algorithm executed in steps 61-69, i.e., taking a plurality of samples to be included in one data block, and obtaining a delta value corresponding to each of the samples. The part storing the data to be represented is very similar to the first part of the algorithm shown in FIG. However, the algorithm of FIG. 4 is the same as the algorithm of FIG. 3 in that the value of the variable MaxDelta is obtained for each data block. However, as shown in blocks 62 and 64, the value stored in the buffer is Instead of the delta value, the actual sample value is stored as it is. Otherwise, steps 61-6 in FIG.
Reference numeral 8 corresponds to steps 51 to 58 in FIG. In step 69, the value of the variable NBits is written to the data block, which is also the same as the algorithm in FIG.
In the following steps 71 to 77, what kind of compressed data is to be determined. In this embodiment, the variable NB
The value of “its” is used as an index value indicating the activity of the signal in the data block. And this variable NB
The threshold value for comparing the values of
Two threshold values are stored, which are shown in steps 70, 72, 74 and 76. Variable NB
The number of samples for actually storing data is determined according to the comparison result between the value of "its" and the four threshold values. That is, first, in step 71, the variable N
If the value of Bits is greater than or equal to the largest threshold value (first threshold value), the routine proceeds to step 71, where all delta values are set as in the lossless compression algorithm of FIG. Is written to the data block. On the other hand, if the value of the variable NBits is less than the first threshold value, the algorithm proceeds to step 72, where the value of the variable NBits is compared to a second threshold value less than the first threshold value. If the value of the variable NBits is equal to or greater than the second threshold value, the delta value (difference value) between every other sample is stored. In this case, the maximum number of digits of the delta value (difference value) may be 2 × MaxDelta, so when writing the difference value,
It is necessary to write up to (NBits + 1) bits from the least significant bit of the difference value. On the other hand, the variable NBit
If the value of s is less than the second threshold value, then the value of the variable NBits is
A comparison is made with a third threshold value that is smaller than the threshold value. If the activity of the signal is relatively high and the value of the variable NBits is equal to or larger than the third threshold value, the difference value between the fourth samples is stored. In this case, from the least significant bit of the difference value, (NBits +
2) try to write up to bits, because
In this case, the number of significant bits of the delta value (difference value)
This is because there is a possibility that the maximum value will be 4 × MaxDelta. If, on the other hand, the value of the variable NBits is still smaller than the third threshold value, then in a step 76 it is compared with the fourth threshold value which is the smallest threshold value. And according to the comparison result,
It is determined whether the delta value (difference value) between every eighth sample is to be stored or whether there is no sample to be stored at all. As shown in step 77, when storing the difference value between every eighth sample, (NBit
Write up to s + 3) bits. Also,
In any of the above cases, when writing the delta value (difference value), one plus / minus sign bit indicating whether the value is a positive value or a negative value is added and written. Once the sample in the buffer has been written to the data block, the routine returns to step 62 and performs the processing for the next data block by repeating the above loop.

The algorithm of FIG. 4 has a constraint on the block size, which means that the block size must be a multiple of "8". When performing decompression (the decompression flowchart is not shown), in the decompression routine,
First, the variable NBi written in the data block
By comparing the value of ts with a plurality of known threshold values, it is possible to determine how many bits each delta value was written when a difference value (delta value) between the sample values was written. judge. Furthermore, the value of the sample that has been discarded without being written may be restored by interpolation at the time of decompression. When the compression results obtained by actually applying the lossy compression method described above to four different types of electrocardiographic signals are analyzed, the compression ratio is about 3.5:
It was in the range of 1 to about 5.5: 1. Further, the analysis revealed that a larger compression ratio was obtained as the threshold values were set to larger values.

Referring now to FIG. Shown in the figure is a flow chart of yet another algorithm, which uses a pacemaker marker marker in the compressed signal.
It is designed to add a channel. The algorithm of FIG. 5 can be applied to both the lossless compression method and the lossy compression method described above. As can be seen from FIG. 5, this algorithm differs from the previously described lossless compression routine in that
That steps 83 and 84 have been added,
In step 90, a signal bit indicating that the signal record is to follow is written.
More specifically, steps 80 to 82 are preparation steps for collecting data of a sample. In the next step 83, the pacemaker determines whether or not a marker has occurred. That is, the pacemaker
Immediately after the generation of the sample clock, a check is performed to determine whether or not a marker has been generated. If the marker has not been generated, the processing accompanying the generation of the new sample clock is executed in the same manner as the algorithm described above, as shown in blocks 85 to 89. On the other hand, if a marker has been generated, step 84
The data is written in. Writing the data for the marker record occurs during the execution of data collection to derive the delta value, and thus occurs prior to writing all the data to the data block at step 90.

The step 84 will be described in further detail. First, when a marker has been generated, one bit (for example, a binary "1") indicating that a "marker record" follows follows. ) To the data block. Subsequently, the value of the variable #Samples (this value is 3 bits or 4 bits) is written. The value of this variable #Samples gives information on where in the data block the marker record was received. by this,
A record is kept of where in the physiological signal the marker occurred. Subsequently, a marker value which is data relating to an actual marker is written.

The algorithm further rewrites one bit (eg, a bit representing a binary "0") at step 90 after performing the entire analysis of the buffer, so that the actual signal record is later Show what follows. This is followed by writing the value of the variable NBits to the data block, and then writing the contents of the buffer to cause the data block to contain data representing the delta value, as in the algorithm described above. Done. In the case of a data block containing a marker record, the head of the data block consists of the data written in step 84, followed by a signal record. The above-mentioned bit indicating that the data block is located is followed by data representing the contents of the data block. On the other hand, for a data block in which no marker has occurred during the period corresponding to the data block, the only difference caused by employing this algorithm is:
Only a single bit is added at the beginning to indicate that the signal record will follow. Therefore, when a marker channel is added to the compressed data as described above, the compression ratio slightly decreases due to the addition of a bit indicating that a signal record follows. The degradation is insignificant compared to the advantage that marker channel data can be added.

Next, FIG. 6A will be described. Shown in the figure is a preferred embodiment for performing non-linear sampling and subsequent A / D conversion. As described above in connection with the embodiment of FIG. 4, if the effective sampling rate is changed in accordance with the instantaneous value of the signal activity, the sample corresponding to the set sampling rate is obtained. Physiological signals can be digitized with fewer samples. For example, when the set sampling rate is 200 / sec, the average value of the effective sampling rate can be set to about 50 / sec to generate high quality ECG and EGM signals. In addition, if it is determined whether each sample should be discarded before the sample is sent to the A / D converter, the power consumption of the battery can be significantly reduced. Further, if the non-linear sampling is performed before the A / D conversion, not only the power consumption can be reduced, but also the amount of data that needs to be stored or processed can be reduced. still,
It should be noted that the term "non-linear" as used herein means non-linear in the time interval of sampling and not the non-linearity of the resolution of the A / D conversion, so that it should not be confused.

In FIG. 6A, a signal source 91
The input signal coming into is directly connected to an A / D converter 96 to digitize it, but the converter 96 is not always enabled. That is, the converter 96 is enabled only when it is determined that the input signal has undergone a sufficiently large fluctuation since the input signal is examined and the sample is finally taken. This will be described in more detail. The sample clock supplied to the input terminal labeled "conversion request" in the figure is not directly connected to the converter 96, and the input signal has a sufficiently large fluctuation. Is supplied to the converter 96 through the gate only when it is determined that should be taken. Whether to take off or take in a certain sample is determined based on a difference value between the input signal and a signal stored in the sample and hold circuit 92, and the difference value is determined by a signal generation circuit. It is compared with the output signal from 94. Sample and hold circuit 9
2 holds the analog value of the newest sample among the samples actually converted, and the held value is input to the window comparison circuit 95 and used as the median value. The output of the sample and hold circuit 92 is supplied to one of three inputs of a window comparison circuit 95 via a switch 93. The second input of the window comparison circuit 95 is Vi
An analog signal input indicated by n. The third input of the window comparison circuit 95 is a signal sent from the signal generation circuit 94. When the conversion of the latest sample has been completed, a "conversion complete" signal is sent out, which is supplied to both the sample and hold circuit 92 and the signal generation circuit 94. When the conversion of one sample is completed, the just-converted value is held in the sample and hold circuit 92,
Further, the signal generation circuit 94 starts transmitting the threshold signal. This threshold signal is a signal which is at a predetermined level at the start of transmission and whose level decreases toward zero over time, and is connected to one of three inputs of the window comparison circuit 95. Have been. This window signal sent by the signal generation circuit 94 is essentially a time-varying threshold signal, and is an analog signal that performs a function similar to the discrete threshold values used in the algorithm of FIG. . The window comparison circuit 95 compares the threshold signal with the difference value of the input signal, and passes the sample clock signal only when the difference value of the input signal is larger than the value of the threshold signal. Supply to conversion start terminal. Therefore, the longer it takes for the value of the output signal of the signal generating circuit 94 to decay to a value smaller than the absolute value of the analog signal when the value of the sample taken most recently is used as the center value, the more it takes Will be forgotten. Then, when the difference value of the input signal exceeds the value of the threshold signal (that is, the value of the input signal protrudes outside the window width), the sample clock signal is taken in and the A / D converter 96 Conversion starts. At the time of this conversion, an input signal input from 91 in the figure is converted at a normal resolution to generate a digital output. At this time, a conversion completion signal is transmitted at the same time. By transmitting the conversion completion signal, the signal generation circuit 94 is reset, and the sample-and-hold circuit 92 is triggered to hold the value of the latest signal. Will be updated to

FIG. 6B is a chart for explaining the function of the above threshold window, and illustrates the relationship between the window envelope and the changing input signal. The time at which the sample was taken is indicated by a square mark in the figure, and when the sample was taken, the signal generation circuit 94 starts transmitting a signal for providing a two-phase window. This signal has positive and negative envelopes, and the input signal is compared to those envelopes. As can be seen from the figure, this window signal has a predetermined value at the start of transmission, and the value attenuates toward zero. This window signal has the center value with the value of the latest analog input signal. However, for a pacemaker operation in which the input signal is a bipolar signal that can take any of positive and negative values, and the basic sampling frequency is a relatively low frequency of, for example, about 200 Hz,
It has been found that setting the center value of this comparison window to 0 volts, rather than to the value of the most recently converted input signal, improves circuit performance. To cope with such a case, an analog signal switch 93 is provided, and by operating this switch 93, the center value of the comparison window can be forcibly set to 0 volt. 6B, the sample clocks indicated by black circles are generated six times after the sample actually taken in the leading part of the curve in the figure. When these sample clocks are generated, , The input signal has changed only slightly compared to the latest sampled value and remains within the window width. Therefore, at the time of generation of these sample clocks, samples are not taken in but are postponed. At the time when the seventh sample clock occurs after the latest sample, the input signal extends outside the window width because the threshold window has been reduced to a very small value. ing. Therefore, a conversion start signal is supplied from the window comparison circuit 95 to the A / D converter 96. The value of the threshold signal is made to be zero after a predetermined number of sample clocks have elapsed, which is at least the minimum sampling time allowed.
Important to ensure that samples are taken at a rate. Further, as is apparent from FIG. 6B, when the activity of the signal is increased, the sampling rate is greatly increased, and the sample of the input signal is actually taken in every time the sample clock is generated. The sampling rate is as follows. This ensures that the actual sampling is a non-linear sampling, i.e. the effective sampling rate is adaptively adjusted according to the relative information content of the physiological signal.

As can be seen from FIG. 6B, of the 47 samples from the first sample to the forty-seventh sample, the number of actually taken samples is reduced to 13, and the reduction rate is Equivalent to 72%. on the other hand,
During the period when the QRS complex is occurring, most of the samples are actually captured, in this case
The rate is effectively as high as the lossless sampling is taking place. Needless to say, the decay rate of the sample to be actually acquired, and thus the magnitude of the distortion of the restored signal, are influenced by the shape (how to reduce) the threshold signal (window signal). Therefore, the shape of the window signal and the timeout time are as follows:
This is a design choice and should be selected or adjusted to suit the characteristics of the signal to be sampled (eg, a particular physiological signal, etc.). In the illustrated embodiment, it is possible to adjust the balance between the magnitude of the signal distortion and the power consumption or the data reduction rate only by modifying the threshold signal generation circuit 94.

In order to exemplify the effect of changing the setting of the threshold signal, respective curves representing two types of threshold signals are shown in FIG. 6C. In each of these threshold signals, the period from the initial value to the decay to zero is set to seven clock cycles. The following table shows, for each of these two curves (Curve 1 and Curve 2), the relationship between the number of samples missed without being captured and the corresponding threshold value.

[0029]

[Table 1]

In a device such as a pacemaker, it is desirable that the threshold window signal can be changed according to the situation. In this case, the change may be set in advance so that the change is made by programming, or the change may be made by performing adjustment based on accumulated data. For example, if the initial value of the threshold signal is reduced during a period in which the amplitude of the signal is small, it is possible to avoid a situation where the distortion is slightly increased due to the small amplitude of the signal. In FIG. 6A, the memory 98 stores various parameter values related to the signal generating circuit 94. For example, these parameter values can be set to the initial value of the threshold signal, the decay rate, and can be forgotten without being taken in. The maximum number of heartbeats allowed. The data of these parameters relating to the threshold signal is supplied to the signal generation circuit 94 via the processor 34. The data of these parameters may be supplied from a programming device that is an extracorporeal device via the transmission / reception circuit 36, or a microprocessor monitors and accumulates the signal data, and is represented by the accumulated signal data. The microprocessor may be programmed to make logical decisions regarding threshold curve optimization based on the signal history. Such adjustment of the threshold curve can be performed in the illustrated embodiment having the analog configuration, and can also be performed in the embodiment having the corresponding digital configuration.

When incorporated in a pacemaker, real-time analog uplink telemetry may be performed by transmitting telemetry directly from the nonlinear sampling A / D converter shown in FIG. 6A. Further, in this case, if the sample value is transferred to the programming device by uplink telemetry, the discarded sample value at the pacemaker side is restored using an interpolation method in the programming device, and the signal is transmitted to the user. May be displayed on the display. With the whole of the above scheme, both the power consumption in the A / D converter and the power consumption in the uplink telemetry circuit are greatly reduced. Although the nonlinear sampling A / D converter according to the preferred embodiment described above is an analog converter, a converter using a digital converter equivalent thereto is also included in the scope of the present invention.

When storing data in a pacemaker or performing further processing on the data using a non-linear sampling circuit, both the amplitude value and the interval value of the sample must be considered. Therefore, when the value (amplitude value) of the input signal is within the window width,
The missed sample clock pulse is provided from the comparison circuit 95 to the counter 97 so that a record of the number of missed samples is kept. In order to accurately store the data of the signal, it is necessary to combine the digital information representing the amplitude value of the sample with the counter information. When the threshold signal is set so that the maximum number of samples that can be forgotten is 7 and the A / D-converted value (amplitude value) is 8 bits, a sample representing one sample The word becomes an 11-bit word, because the 8-bit amplitude value is appended with a 3-bit number representing the elapsed time (ie, the number of sample clocks that have been dropped since the last sample was taken). This is because that. Furthermore, this means that taking in one sample adds three bits of overhead, but overall the memory required to store the signal is small and the battery is low. The power consumption will also be reduced.

As described above, the stream of 11-bit sample words generated by the circuit of FIG. 6A includes a lossless compression scheme for compressing amplitude values and a stream for compressing elapsed time values. Further compression can be achieved by using the fixed Huffman coding system in combination. In this regard, various physiological signals were sampled using the above-described nonlinear sampling method, and the distribution of the number of samples that were not taken in at the time of sampling (ie, elapsed time) was examined. For example, it has been found that there is no strong correlation or similarity in the distribution between the signals. This means that it is difficult to compress the data representing the number of samples that have been sent off without taking in the data as it is. However, when examining the number difference (delta value) between the number of samples deferred before capturing a certain sample and the number of samples deferred before capturing the next sample, even if the original signal is different, As for the delta value of the number of samples forgotten, it was found that there was a strong correlation between the signals. This means that the delta value (the difference in the number of samples that have been forgotten) is a compressible parameter. However, the disadvantage of using this parameter is that the range of possible values for this parameter is "-7".
To "+7". Therefore, when storing a parameter that directly represents the number of samples that have been forgotten, only three bits are required, whereas storing this delta value requires four bits. It is. However, if the remainder (modulo 8) obtained by dividing the delta value by “8” is used, the range of the value can be returned from “0” to “7” again. Individual values representing the number can be reproduced. If this value is changed, the negative delta value is converted to a positive value according to the calculation formula "8- (absolute value of the delta value)". Therefore, for example, if the number of samples
When the number changes from three to three (that is, delta value = -3)
And the case where the delta value is +5, the same result is obtained. In the latter case, the following expression is obtained. (6 + 5) MOD8 = 11 MOD8 = 3

It was found that a strong correlation still exists in the value of the finally derived parameter “remainder (modulo 8) obtained by dividing the delta value (the difference in the number of samples sent off)” by “8”. ing. Therefore, based on this distribution, it is possible to create a fixed Huffman code that more efficiently represents a parameter of elapsed time (that is, the number of samples that have been forgotten). There are various specific methods for creating a Huffman code tree based on an arbitrary probability distribution, and they are described in various documents. If a physiological signal is sampled using a threshold signal as shown in Table 1 and FIG.
From the result, the delta value (difference in the number of samples sent off)
The distribution of the value of the remainder (modulo 8) obtained by dividing by “8” may be obtained, and a Huffman coding tree can be created based on the distribution. The resulting code is, for example, as shown in Table 2 below. Take the value as shown.

[0035]

[Table 2]

By using the Huffman coding shown in Table 2, the first threshold set value is about 17.
A reduction in the number of bits of 5% was achieved, and a reduction in the number of bits of about 14.5% was achieved for the second threshold setting. Since a fixed Huffman code is used, this coding is performed by reading a code corresponding to each parameter from a very small look-up table, and requires little processing.

Next, FIG. 7 will be described. FIG. 7 is a flowchart showing an entire compression algorithm for processing a signal sampled by the non-linear sampling method. In step 101, the value of the variable LastSample representing the value of the latest sample taken is set to “0”.
, And a variable LastSkip representing the number of samples deferred until the latest sample is captured.
Set the value of ped to "0". Steps 103 to 10
Reference numeral 7 denotes a step for assembling data sampled by the non-linear sampling method to form a data block. This data block is formed in the same manner as in the lossless compression method described above.
In step 106, a value of a variable MaxDelta indicating the maximum value of the absolute value of the delta value is set. Step 103 represents an operation for nonlinear sampling performed by the circuit shown in FIG. 6A. Step 104 shows the main differences between the algorithm of FIG. 7 and the conventional method which is the premise of the present invention shown in FIG. 3, that is, the algorithm of FIG. Delta value ADDelta representing the difference between the variables #Sam
amplitude buffer ABuf [# corresponding to the value of ples
Samples], and a delta value TDelta representing a difference between elapsed time values (more precisely, a remainder (modulo 8) obtained by dividing the delta value by “8”) in a variable #
Time value buffer TBu corresponding to the value of Samples
f [#Samples]. Steps 108 to 112 are steps of the compression processing. In these steps, first, the variable MaxDe
It determines the number of significant bits of the value of lta, and then, in step 109, a variable NB representing the number of significant bits.
Write the value of "its" as a block header. Subsequently, in steps 110, 111, and 112, data representing the elapsed time value of the sample and data representing the amplitude value of the sample are all written in the data block. More specifically, the value of the variable #Samples is a variable Buffer
In step 110, it is determined whether or not the value is equal to the value of Size. In step 111, the value of the variable #Samples is incremented by "1" until they become equal. To explain in more detail, first,
A Huffman code representing a remainder (modulo 8) obtained by dividing the delta value (the difference in the number of samples that have been forgotten) by “8” is written, and then the least significant bit of the bits of the delta value representing the difference between the amplitude values To NBits bits (plus one sign bit). Subsequently, the value of the variable #Samples is changed to the variable BufferSi.
Compared with the value of ze, the value of variable #Samples is set to “1”.
After the writing of all the elapsed time value data and amplitude value data corresponding to the current data block is completed by repeating
The routine returns from step 110 to step 102 to begin work on the next data block.

To retrieve information from a data block, first read the value of the variable NBits. Subsequently, all of a plurality of sample values each including elapsed time value information (elapsed time value data) and amplitude value information (amplitude value data) are extracted. When extracting elapsed time value information, each time a bit of data representing an elapsed time value is read, it is checked against the Huffman code tree, and this operation is performed on any leaf node in the tree. Until it reaches. The content of the leaf node is a remainder (modu) obtained by dividing the delta value (the difference in the number of samples that have been forgotten) by “8”.
lo8) is a decompressed value. The value obtained in this way is added to the elapsed time value of the preceding sample, and the sum is divided by “8” to obtain a remainder (modulo).
Taking 8) gives the elapsed time value of the current sample. To extract the amplitude value information, NBi is calculated from the sign bit and the least significant bit of the current delta value.
Up to ts bits are read, and the read value is added to the amplitude value of the preceding sample. Finally, the interpolation is applied to the values of the samples taken out as described above, and the discarded samples are restored to reproduce the entire original signal. Using this method, the data sent by the non-linear sampling A / D converter can be further compressed by about 25%, and the overall compression ratio is 4: 1 to 5: 1 with respect to the original signal.

In the above description, the present invention has been disclosed in the case of handling a physiological signal such as an electrocardiographic signal. However, the present invention is not limited to the case of handling such a physiological signal. It can be used to process various signals such as sensor signals in application fields that are significantly different from this, such as meteorology and seismology. In particular, in a battery-powered device environment, the system of the present invention is very useful in achieving very high compression ratios with relatively little information loss. According to the non-linear sampling A / D converter according to the present invention, not only the power consumption in the digitization process is reduced, but also the power consumption in other processes connected to the digitization process in a series of processing procedures. Reduced.
Furthermore, the non-linear sampling according to the present invention can also be executed in the embodiment of the digital configuration, and is not limited to the embodiment of the analog configuration disclosed above.

[Brief description of the drawings]

FIG. 1A is a schematic diagram showing two examples of devices to which the present invention can be applied, an implantable device such as a pacemaker and a battery-powered holter-type device that is worn on the body surface of a driven patient. B is a simplified block diagram showing the major components of the implantable device or body surface mounted device shown in A.

FIG. 2 is a simplified flowchart showing the major steps performed in compressing, transferring, and decompressing data according to the present invention.

FIG. 3 is a flowchart showing a specific example of a data lossless compression method according to a conventional example.

FIG. 4 is a flow chart of a lossy compression algorithm for performing non-linear sampling as a function of a predetermined threshold value.

FIG. 5 is a flow chart of an algorithm for appending pacemaker marker channel data to a data stream that is compressed in accordance with the present invention.

6A is a circuit diagram of an embodiment of an analog configuration for non-linear sampling and A / D conversion, FIG. 6B is a chart illustrating the operation of the non-linear A / D converter of A, and FIG. 4 is a graph illustrating two types of threshold value curves that can be used for a nonlinear A / D converter.

FIG. 7 is a flow chart of an algorithm according to the present invention that performs non-linear sampled data compression to obtain the advantages of both non-linear sampling and data compression.

[Explanation of symbols]

 Reference Signs List 15 battery 30 implantable device 32 body surface-mounted device 34 processor (microprocessor) 35 memory 92 sample and hold circuit 94 signal generation circuit 95 window comparison circuit 96 A / D converter 98 memory

Continuing on the front page (72) Inventor Korn Weyant Nüll-6431 The Netherlands Hoehbroek, Hommelweerweg 217

Claims (30)

[Claims] 1. A system for sampling an analog signal in a non-linear sampling manner to generate compressed digital data representing the analog signal, comprising: a clock signal source for transmitting a sample clock signal at a predetermined clock rate; Sampling means for generating digital data representing the analog signal by sampling the analog signal when a clock signal is generated; holding means for holding each sample value of the analog signal; sampling of the analog signal Threshold means for generating a difference threshold signal representing a dynamically decreasing difference with respect to the held sample value each time the sampled signal is transmitted after the sampling is performed. , A sample acquisition permitting means for judging whether a difference between the analog signal and the held sample value is larger than the threshold value and transmitting a sample acquisition permission signal if the difference is larger than the threshold value; And only when the sample capture permission signal is transmitted, each time the sample clock signal is transmitted, enables the sampling means to cause the sampling means to sample the analog signal,
Thereby sampling off at a non-linear sampling rate where some of the samples are reduced according to the time difference of the analog signal and the way in which the threshold value is reduced, without taking some of the samples. Enabling means to be performed. 2. The method according to claim 1, further comprising: determining a number of samples that have been skipped without being acquired after the sampling is performed and before performing the next sampling; and generating digital data representing the number of samples that have been skipped. The system of claim 1, further comprising means. 3. The apparatus according to claim 2, further comprising storage means for storing digital data representing the analog signal and digital data representing the elapsed time in association with each other for each sample. system. 4. The signal processing apparatus according to claim 1, wherein said threshold means includes a signal generating means, and a signal generated by said signal generating means has a predetermined number of samples after the amplitude thereof has decreased from a predetermined initial amplitude. The system of claim 1, wherein the signal decreases to substantially zero within a predetermined time corresponding to a time until a clock is generated. 5. A compression system adapted to adaptively adjust an effective sampling rate of an analog signal as a function of the variability of the analog signal, wherein the sampling means samples the analog signal at a predetermined sampling rate. Block means for holding a predetermined number of samples obtained by sampling; delta means for generating a delta value representing a difference between the value of each held sample and the value of a sample held prior to the sample Maximum value means for obtaining a maximum delta value in the block for each of the blocks; threshold means for setting a plurality of threshold values corresponding to a predetermined number of different delta values; Is compared with the threshold value of Comparing means for maintaining some of the delta values and discarding others, and for writing, for each of the blocks, compressed data consisting of the maximum delta value and the maintained delta value. Compression means comprising: 6. The compression system according to claim 5, wherein said writing means includes telemetry means for transmitting said compressed data directly by telemetry. 7. The compression system according to claim 5, wherein said writing means includes storage means for storing said compressed data. 8. The threshold means comprises means for storing a plurality of discrete threshold values, and wherein the comparing means determines that the maximum delta value is between any two of the plurality of threshold values. 6. The compression system according to claim 5, further comprising means for judging whether or not the delta value exists, and discarding the delta value as a function of the judgment result. 9. The writing means includes bit number determining means, and the bit number determining means determines the number of bits to be written when writing each of the maintained delta values as a function of the determination result. 9. The compression system according to claim 8, wherein said means is a determining means. 10. The compression system built into a pacemaker and further comprising means for writing pacemaker marker data into the compressed data if a marker event occurs during the block. The compression system according to claim 5, wherein 11. Pacing means for pacing a patient, detecting means for detecting an analog signal as a physiological signal, and data means for generating compressed digital data representing the analog signal from the analog signal. An implantable pacemaker system comprising: a data rate measuring means for deciding a variability index value of the analog signal; and a non-linear means for extracting sample data from the analog signal at a non-linear sampling rate. Wherein said non-linear means includes non-linear sampling rate determining means for determining said non-linear sampling rate as a function of said variability index value; and data representing said non-linear sampling rate and said extracted samples. Writing means for writing data Implantable pacemaker system, characterized in that. 12. The non-linear means includes clock means for generating a clock signal, sampling means for sampling the analog signal each time the clock signal is generated, and converter means for converting each sample into digital data. And sample sending-off means. The sample sending-off means causes the sampling means to perform a sampling operation in response to a portion of the analog signal containing only a small change with respect to the variability index value. 12. The implantable pacemaker according to claim 11, further comprising means for causing the sample to be skipped without delay, thereby reducing the number of sampling operations to less than the number of occurrences of the sample clock.
system. 13. The implantable type according to claim 12, further comprising: means for generating data representing the number of samples sent off without being taken in; and means for compressing the number of samples to be sent off for compressing the data. Pacemaker system. 14. The number-of-offset-samples number compressing means,
A delta means for generating offset delta data representing the difference between the number of samples deferred before a sample is acquired and the number of samples deferred until the next sample is acquired. 13. The implantable pacemaker system according to claim 13. 15. The implantable type according to claim 13, further comprising lossless compression means for compressing said sample digital data, and coding means for coding said delay delta data. Pacemaker system. 16. A method for generating compressed data representing an analog signal, comprising: setting threshold data associated with a predicted value of the analog signal; and sampling the analog signal at a non-linear effective sampling rate. Generate data,
At this time, the analog signal is compared with the threshold data, the nonlinear sampling rate is determined based on the comparison result, and an index value representing the nonlinear sampling rate corresponding to each sample is derived. Converting the index value into rate data, compressing the sample data by a lossless compression method, and writing the rate data and the sample data. 17. The method of claim 16, wherein analog threshold data is set and the non-linear sampling rate is determined before sampling the analog signal. 18. A method according to claim 1, wherein a plurality of discrete threshold data points are set, and the analog signal is sampled at a constant sampling rate first.
Extracting data, comparing the first sample data to the plurality of threshold data points, and sampling the first sample data at a non-linear sampling rate as a function of the result of the comparison. Item 17. The method according to Item 16. 19. A method for sampling at a predetermined clock rate.
A non-acquisition sample or a non-acquisition sample at each occurrence of the sample clock, and determining the number of samples skipped without acquisition by determining the number of non-acquisition samples. 17. The method according to claim 16, wherein said index value representing a rate is derived. 20. The method according to claim 10, wherein a difference between the number of samples deferred until a certain sample is captured and the number of samples deferred before the next sample is captured is calculated and the number difference is encoded. The method of claim 19, wherein the data is compressed. 21. Signal means for generating a physiological signal as an analog signal; and data means for generating a compressed digital data stream representative of the analog signal, wherein the data means comprises the analog signal at a non-linear sample interval. Non-linear means for taking in a sample of the sample, first compression means for generating compressed data representing the sample, and second compression means for generating compressed data representing a sample interval of the sample, Generate a data stream representing an analog signal,
An implantable medical device system, comprising: 22. The implantable medical device system is a pacemaker, the pacemaker having marker means for capturing marker channel data, and inserting means for inserting the marker channel data into the data stream. 22. The implantable medical device system according to claim 21, comprising: 23. The implantable medical device according to claim 21, wherein said non-linear means includes an analog non-linear sampling A / D converter for determining a sampling rate before performing A / D conversion. Device system. 24. A / D sampling means for generating a plurality of digital samples representative of the analog signal, and one of the plurality of digital samples as a function of a rate of variation of the analog signal. 22. The implantable medical device system according to claim 21, further comprising sample sending means for sending off the sample without taking in the sample. 25. The non-linear means comprises threshold signal means for generating a threshold signal, wherein the threshold signal means starts generating the threshold signal each time the sample is taken. 22. The implantable medical device system according to claim 21. 26. The apparatus further comprising: clock means for generating a clock signal; and comparison means operable to compare the threshold signal and the analog signal each time the clock signal is generated. Item 29. The implantable medical device system according to Item 25. 27. The implantable medical device system according to claim 25, further comprising means for adjusting the threshold signal as a function of signal history. 28. The implantable medical device system according to claim 25, further comprising program means for adjusting the threshold signal as a function of programmed data. 29. The apparatus according to claim 2, wherein the threshold signal has an initial value and a length, and further comprises adjusting means for adjusting at least one of the initial value and the length.
5. The implantable medical device system of claim 5. 30. The plant according to claim 29, wherein said adjusting means includes storing means for storing data representing said analog signal and adjusting said threshold signal as a function of said stored data. Embedded medical device system.